The ever-evolving nature of Gen AI - how can you keep pace?

The ever-evolving nature of Gen AI - how can you keep pace?

By Dele Ogunjumelo, Principal Engineer and Head of Machine Learning Guild at Mobica

Generative AI (Gen AI) is a truly disruptive force.

One impact study from Cognizant predicts 90% of jobs will be affected in some way by the technology.

But just what is Gen AI?

Gen AI is a subset of artificial intelligence. It creates algorithms and models that generate new content, such as text, photos, code, videos, 3D renderings, music and more. By using deep learning techniques and transformer architectures, it can learn patterns and structures within data to produce this new content, without explicit programming.

Every sector will feel the impact. From increased productivity to new functionalities, Gen AI represents a step change in what’s possible. Coupled with low barriers to entry for users - courtesy of intuitive and interactive interfaces such as gesture and voice recognition technology - every business and workforce can reap the benefits of this innovation.

Yet the very nature of Gen AI means it’s constantly evolving. How, then, can companies keep up?

Let’s dive into the world of Gen AI. Here, I’ll explain some of the opportunities to push ahead, consider the industry’s response to this innovation to date, and make clear how Mobica is keeping pace.

gen-ai

Reshaping software expectations

Gen AI is proving its value already, across a wide range of industries.

In the automotive sphere, Gen AI is being used for everything from traffic route optimisation to recommendations for hotels, restaurants and flights. By using a speech interface, in-vehicle assistants can make proactive recommendations based on what they know about the driver and data collected by the vehicle.

In the industrial space, there are so many sensors in devices and machinery that collect data and then analyse it. This intends to predict potential risks and identify any possible failures, as well as understand how safety, quality and efficiency can be improved. But, of course, the Industrial Internet of Things has been heading this way for some time. Gen AI goes one step further.

With access to vast amounts of data, which could include documentation on international standards, data or parameters from a product under design could be sent to a Gen AI system to make sure it complies with these standards. The extra step that Gen AI can then take is corrective action, to automatically get a product where it needs to be.

The healthcare industry is also set to gain from Gen AI. For instance, with around 330,000 new cases of skin cancer every year1, Gen AI could offer a path for improved and more accurate diagnosis.

As part of an internal project, we’ve implemented a CNN machine-learning model that can help detect cancer in the skin. Gen AI is then used to provide personalised treatment recommendations based on skin cancer classification and other data such as sex and age.

All of these advances are having an impact on the semiconductor sector, too. As you can imagine, to support the aforementioned strides in these industries, silicon businesses are having to develop hardware architectures that can improve the performance of Gen AI solutions. A variety of hardware and architectures are being proposed for Gen AI, which aim to strike a balance between scalability, flexibility, and efficiency. Techniques to reduce complexity are being explored in order to reduce the size of generative models without compromising their performance significantly.

Also, work is being done on the design of circuits and systems for efficient Gen AI implementation. This includes designing the architecture for Gen AI, designing circuits for ASICs and FPGAs, and developing embedded software for multi-core processors, GPUs, DSPs and machine-learning accelerators.

There is clearly a need for advanced Gen AI accelerators, since circuits and systems, which are limited by fixed computational resources, find it difficult to keep up with the rapid expansion of Gen AI models.

With speech prompts and gesture recognition giving the user a more interactive interface, it makes this technology so easy to embrace, Gen AI can be put to work in all kinds of use cases - the opportunities are endless. However, this leads us to the next question: where should businesses focus their efforts?

speech-prompt

Areas of untapped potential for gen AI

With so much to gain from Gen AI, what use cases would companies be wise to capitalise on? Two key areas to consider are how Gen AI can help with the development of your business’ products and services, and assist with internal processes too.

For instance, when developing AI models, a lack of access to appropriate data can be a common challenge. From e-commerce companies trying to build recommendation systems, who are held back by other businesses unwilling to share their data, to healthcare organisations that are undertaking medical research, who have limited data due to confidentiality reasons, Gen AI has the answer - synthetic data generation.

This uses a prompt with some kind of context to artificially generate data that mimics real-world data, removing the challenge of not having access to this information. Whether you need to have the right data to develop a solution, or when looking to prove the viability of a future business case, synthetic data generation has a powerful role to play.

Another example might be when an internal team is looking to test a system under different schemes. This might be an automotive team wanting to create a range of driving scenarios - different times, dates, weather conditions and driving environments. Gen AI can generate these simulations and measure how a vehicle performs, automatically identifying or even implementing recommendations for improvements. Ultimately, Gen AI can help you prove your competitive advantage when compared with other brands.

automotive-team

Overcoming Gen AI obstacles

While gen AI presents an incredible opportunity for companies, there are limitations. It requires large amounts of data, and outputs may not be accurate if the input data isn’t reliable. Indeed, Gen AI is also not a silver bullet - for many initiatives, it might not be needed.

But for projects where Gen AI could drive new efficiencies, decision-makers need to be aware of possible obstacles they’ll need to navigate.

Firstly, with this evolving technology constantly advancing, having the right tech expertise to deploy and develop these solutions is absolutely critical. One recent report from the Enterprise Strategy Group found 97% of organisations with Gen AI projects consider it a top-five priority. And yet 61% of respondents also said most users don’t know how to capitalise on Gen AI, with 51% also reporting a lack of employees with Gen AI expertise. As a result, there’s a need to invest in continual upskilling programmes too.

 

Another common challenge is a business’ infrastructure readiness. Do you have a scalable and flexible infrastructure that can support growing Gen AI demands? Many opt for the cloud or a hybrid approach, but even the cloud has its limitations - would a localised network serve you better, such as PrivateGPT? This can not only help you reduce the cost of accessing cloud services, but also solve the issue of security, as confidential information is not sent to the cloud.

There’s also the issue of privacy and security. For Gen AI applications, companies need to implement a privacy and security framework for the data they’re dealing with, and then find a way to enforce this. Third-party information must be covered as well.

Finally, because of the pace with which Gen AI is accelerating, there’s the very real risk of quickly being left behind. Technology can become obsolete over time which is why it pays to tap into a community of experts, who are constantly keeping an eye on the pulse of Gen AI.

Preparing for what’s next

The realm of Gen AI is accelerating fast. You can see this across industry, with the increasing number of Gen AI-based open source projects available. But the key to success will always be a custom approach to Gen AI solutions, enabling companies to achieve competitive differentiation.

At Mobica, we're dedicated to nurturing a robust community of Gen AI experts. We collaborate with our parent company, Cognizant, and provide our expertise to Qualcomm and other major brands and businesses on embedded Gen AI solutions for devices and platforms. Our team is passionate about exploring projects internally, so we’re able to confidently advise on how Gen AI can address key industry challenges and help businesses maintain a competitive edge. This comprehensive approach means we can stay at the forefront of technological advancements, and drive innovation in the field.

To realise the full potential of Gen AI, Mobica is striving to evolve at the same pace as this ground-breaking technology. Only then can we unlock its true power.

To learn more about how Mobica can accelerate your Gen AI projects, please get in touch.


1 World Cancer Research Fund International, ‘Skin cancer statistics’, https://www.wcrf.org/cancer-trends/skin-cancer-statistics/ 

2 Hitachi Vantara, ‘Lack of governance, infrastructure readiness and IT talent leading to enterprise Gen AI struggles: New report’, https://www.hitachivantara.com/en-us/news/gl240709